Don't leave debug locations for constants sitting on the builder indefinitely
Because constants are currently emitted *before* the prologue, leaving the debug location on the IRBuilder spills onto other instructions in the prologue and messes up both line numbers as well as the point LLVM chooses to be the prologue end.
Example LLVM IR (irrelevant IR elided):
Before:
```
define internal { i64, i64 } `@_ZN3tmp3Foo18var_return_opt_try17he02116165b0fc08cE(ptr` align 8 %self) !dbg !347 { start:
%self.dbg.spill = alloca [8 x i8], align 8
%_0 = alloca [16 x i8], align 8
%residual.dbg.spill = alloca [0 x i8], align 1
#dbg_declare(ptr %residual.dbg.spill, !353, !DIExpression(), !357)
store ptr %self, ptr %self.dbg.spill, align 8, !dbg !357
#dbg_declare(ptr %self.dbg.spill, !350, !DIExpression(), !358)
```
After:
```
define internal { i64, i64 } `@_ZN3tmp3Foo18var_return_opt_try17h00b17d08874ddd90E(ptr` align 8 %self) !dbg !347 { start:
%self.dbg.spill = alloca [8 x i8], align 8
%_0 = alloca [16 x i8], align 8
%residual.dbg.spill = alloca [0 x i8], align 1
#dbg_declare(ptr %residual.dbg.spill, !353, !DIExpression(), !357)
store ptr %self, ptr %self.dbg.spill, align 8
#dbg_declare(ptr %self.dbg.spill, !350, !DIExpression(), !358)
```
Note in particular how !357 from %residual.dbg.spill's dbg_declare no longer falls through onto the store to %self.dbg.spill. This fixes argument values at entry when the constant is a ZST (e.g. `<Option as Try>::Residual`). This fixes#130003 (but note that it does *not* fix issues with argument values and non-ZST constants, which emit their own stores that have debug info on them, like #128945).
r? `@michaelwoerister`
Use more correct handling of lint attributes
The previous analysis was top-down, and worked on a single file (expanding macros). The new analysis is bottom-up, starting from the diagnostics and climbing up the syntax and module tree.
While this is more efficient (and in fact, efficiency was the motivating reason to work on this), unfortunately the code was already fast enough. But luckily, it also fixes a correctness problem: outline parent modules' attributes were not respected for the previous analysis. Case lints specifically did their own analysis to accommodate that, but it was limited to only them. The new analysis works on all kinds of lints, present and future.
It was basically impossible to fix the old analysis without rewriting it because navigating the module hierarchy must come bottom-up, and if we already have a bottom-up analysis (including syntax analysis because modules can be nested in other syntax elements, including macros), it makes sense to use only this kind of analysis.
Few other bugs (not fundamental to the previous analysis) are also fixed, e.g. overwriting of lint levels (i.e. `#[allow(lint)] mod foo { #[warn(lint)] mod bar; }`.
After this PR is merged I intend to work on an editor command that does workspace-wide diagnostics analysis (that is, `rust-analyzer diagnostics` but from your editor and without having to spawn a new process, which will have to analyze the workspace from scratch). This can be useful to users who do not want to enable check on save because of its overhead, but want to see workspace wide diagnostics from r-a (or to maintainers of rust-analyzer).
Closes#18086.
Closes#18081.
Fixes#18056.
The previous analysis was top-down, and worked on a single file (expanding macros). The new analysis is bottom-up, starting from the diagnostics and climbing up the syntax and module tree.
While this is more efficient (and in fact, efficiency was the motivating reason to work on this), unfortunately the code was already fast enough. But luckily, it also fixes a correctness problem: outline parent modules' attributes were not respected for the previous analysis. Case lints specifically did their own analysis to accommodate that, but it was limited to only them. The new analysis works on all kinds of lints, present and future.
It was basically impossible to fix the old analysis without rewriting it because navigating the module hierarchy must come bottom-up, and if we already have a bottom-up analysis (including syntax analysis because modules can be nested in other syntax elements, including macros), it makes sense to use only this kind of analysis.
Few other bugs (not fundamental ti the previous analysis) are also fixed, e.g. overwriting of lint levels (i.e. `#[allow(lint)] mod foo { #[warn(lint)] mod bar; }`.
feat: generate names for tuple-struct in add-missing-match-arms
fix#18034.
This PR includes the following enhancement:
- Introduced a `NameGenerator` in `suggest_name`, which implements an automatic renaming algorithm to avoid name conflicts. Here are a few examples:
```rust
let mut generator = NameGenerator::new();
assert_eq!(generator.suggest_name("a"), "a");
assert_eq!(generator.suggest_name("a"), "a1");
assert_eq!(generator.suggest_name("a"), "a2");
assert_eq!(generator.suggest_name("b"), "b");
assert_eq!(generator.suggest_name("b"), "b1");
assert_eq!(generator.suggest_name("b2"), "b2");
assert_eq!(generator.suggest_name("b"), "b3");
assert_eq!(generator.suggest_name("b"), "b4");
assert_eq!(generator.suggest_name("b3"), "b5");
```
- Updated existing testcases in ide-assists for the new `NameGenerator` (only modified generated names).
- Generate names for tuple structs instead of using wildcard patterns in `add-missing-match-arms`.
fix: Faulty notifications should not bring down the server
Fixes https://github.com/rust-lang/rust-analyzer/issues/18055, if a client sends us an unregistered document path in a did save notification it would force us to exit the thread. That is obviously not great behavior, we should be more fallible here
feat: render patterns in params for hovering
Fix#17858
This PR introduces an option to [hir-def/src/body/pretty.rs](08c7bbc2db/crates/hir-def/src/body/pretty.rs) to render the result as a single line, which is then reused for rendering patterns in parameters for hovering.
assist: ensure `replace_qualified_name_with_use` applies to the first path segment
This change helps a bit with the discoverability of `replace_qualified_name_with_use`. Specifically, it ensures that a cursor on the first path segment (e.g., `$0std::fmt::Debug`, where `$0` is the cursor) would result in an import along the lines of `use std::fmt;` and `fmt::Debug;` at the usage sites.
internal: Better testing infra for ratoml
This PR makes some improvements on how we test configs that come from `rust-analyzer.toml` files.
It was primarily used to solve #18021 but along the way I could not really determine the cause of the said issue which makes me think that it may not be related to the changes that I made earlier to the ratoml infra. In either way `custom_snippets` are now made `global` because we still don't have a tree that maps a `SourceRootId` to a set of `Snippet`s.
fix: Fix `inline_const_as_literal` error when the number >= 10
## Description
### The Bug
This PR fixes a small bug in the IDE assistence (`inline_const_as_literal`). When the being-inlined constant is a number and it is greater than or equal to 10, the assistence inserts unexpected string `(0x...)` after the number itself. A simple example is followed:
Current `inline_const_as_literal` changes
```rs
const A: usize = 16;
fn f() -> usize {
A // inline the constant
}
```
into
```rs
const A: usize = 16;
fn f() -> usize {
16 (0x10)
}
```
The bug originates from #14925 & #15306 . #14925 added some unittests, but it just tested the number-inlining behavior when the number is `0`.
50882fbfa2/crates/ide-assists/src/handlers/inline_const_as_literal.rs (L124-L138)
And #15306 modified the behavior of `Const::render_eval` and added the `(0x...)` part after the number (if the number >= `10`). Because of insufficient unittests in #14925, changes about `Const::render_eval` in #15306 introduced this bug with no CI failure.
### The Fix
I think `Const::render_eval` is intended for user-facing value displaying (e.g. hover) and not designed for `inline_const_as_literal`. To fix the bug, I defined a new function named `Const::eval`, which evaluates the value itself faithfully and simply and does nothing else.
## Thanks
Thanks `@roife` for your kind help. Your guidance helped me better understand the code.
feat: Automatically add semicolon when completing unit-returning functions
But provide a config to suppress that.
I didn't check whether we are in statement expression position, because this is hard in completion (due to the natural incompleteness of source code when completion is invoked), and anyway using function returning unit as an argument to something seems... dubious.
Fixes#17263.
assist: ensure `replace_qualified_name_with_use` applies to the first path segment
This change helps a bit with the discoverability of `replace_qualified_name_with_use`. Specifically, it ensures that a cursor on the first path segment (e.g., `$0std::fmt::Debug`, where `$0` is the cursor) would result in an import along the lines of `use std::fmt;` and `fmt::Debug;` at the usage sites.
Skip checks for cast to dyn traits
It seems that chalk fails to solve some obvious goals when there are some recursiveness in trait environments.
And it doesn't support trait upcasting yet. rust-lang/chalk#796
This PR just skips for casting into types containing `dyn Trait` to prevent false positive diagnostics like #18047 and #18083
Enumerate lint expectations using AttrId
This PR implements the idea I outlined in https://github.com/rust-lang/rust/issues/127884#issuecomment-2240338547
We can uniquely identify a lint expectation `#[expect(lint0, lint1...)]` using the `AttrId` and the index of the lint inside the attribute. This PR uses this property in `check_expectations`.
In addition, this PR stops stashing expected diagnostics to wait for the unstable -> stable `LintExpectationId` mapping: if the lint is emitted with an unstable attribute, it must have been emitted by an `eval_always` query (like inside the resolver), so won't be loaded from cache. Decoding an `AttrId` from the on-disk cache ICEs, so we have no risk of accidentally checking an expectation.
Fixes https://github.com/rust-lang/rust/issues/127884
cc `@xFrednet`
fix: Correctly escape strings in our quote macro
This is a small change, but it was the cause of 90% of the errors in `rust-analyzer diagnostics .` 🫢 (because this worked incorrectly with `stringify!()`, which means every `quote!()` (the original one) quoting a string also didn't work).
With this change and #18085 together, all remaining errors are type errors.
This may mean we can enable more errors, but this is out of scope for this PR.
This is a small change, but it was the cause of 90% of the errors in `rust-analyzer diagnostics .` 🫢
With this change and #18085 together, all remaining errors are type errors.
This may mean we can enable more errors, but this is out of scope for this PR.
Also emit `missing_docs` lint with `--test` to fulfil expectations
This PR removes the "test harness" suppression of the `missing_docs` lint to be able to fulfil `#[expect]` (expectations) as it is now "relevant".
I think the goal was to maybe avoid false-positive while linting on public items under `#[cfg(test)]` but with effective visibility we should no longer have any false-positive.
Another possibility would be to query the lint level and only emit the lint if it's of expect level, but that is even more hacky.
Fixes https://github.com/rust-lang/rust/issues/130021
try-job: x86_64-gnu-aux
internal: Add preliminary `SyntaxEditor` functionality
Related to #15710
Implements a `SyntaxEditor` interface to abstract over the details of modifying syntax trees, to both simplify creating new code fixes and code actions, as well as start on the path of getting rid of mutable syntax nodes.
`SyntaxEditor` relies on `SyntaxMappingBuilder`s to feed in the correct information to map AST nodes created by `make` constructors, as `make` constructors do not guarantee that node identity is preserved. This is to paper over the fact that `make` constructors simply re-parse text input instead of building AST nodes from the ground up and re-using the provided syntax nodes.
`SyntaxAnnotation`s are used to find where syntax elements have ended up after edits are applied. This is primarily useful for the `add_{placeholder,tabstop}` set of methods on `SourceChangeBuilder`, as that currently relies on the nodes provided being in the final syntax tree.
Eventually, the goal should be to move this into the `rowan` crate when we move away from mutable syntax nodes, but for now it'll stay in the `syntax` crate.
---
Closes#14921 as `SyntaxEditor` ensures that all replace changes are disjoint
Closes#9649 by implementing `SyntaxAnnotation`s
rustdoc rfc#3662 changes under unstable flags
* All new functionality is under unstable options
* Adds `--merge=shared|none|finalize` flags
* Adds `--parts-out-dir=<crate specific directory>` for `--merge=none`
to write cross-crate info file for a single crate
* Adds `--include-parts-dir=<previously specified directory>` for
`--merge=finalize` to write cross-crate info files
* `tests/rustdoc/` tests for the new flags
Distribute rustc_codegen_cranelift for Windows
With support for raw-dylib recently added to cg_clif, and inline assembly support working on Windows for quite a while now, all blockers for distributing cg_clif on Windows that I mentioned in https://github.com/rust-lang/rust/pull/81746#issuecomment-1774099637 are fixed now.
feat: better name suggestions for fn
fix#17631.
Better name suggestions for fn-calls / method-calls in the form of `from()`, `from_xxx()`, `into()`, etc.
Correctly handle stability of `#[diagnostic]` attributes
This commit changes the way we treat the stability of attributes in the
`#[diagnostic]` namespace. Instead of relaying on ad-hoc checks to
ensure at call side that a certain attribute is really usable at that
location it centralises the logic to one place. For diagnostic
attributes comming from other crates it just skips serializing
attributes that are not stable and that do not have the corresponding
feature enabled. For attributes from the current crate we can just use
the feature information provided by `TyCtx`.
r? `@compiler-errors`
But provide a config to suppress that.
I didn't check whether we are in statement expression position, because this is hard in completion (due to the natural incompleteness of source code when completion is invoked), and anyway using function returning unit as an argument to something seems... dubious.
